Langston, Psychology of Language, Notes 9 -- Origins of Language
Note:  These notes are in development.  Eventually, a new and nicer version will go here. The powerpoint for notes 9 is more current.

I.  Goals.
A.  Some big debates.
B.  Pinker's arguments.
II.  The big debates.  Some of the issues that get psycholinguists exercised:
A.  Where does language come from?  Are you born with a language organ (in the same way that you're born with a liver), or do you learn your language just like you learn everything else?  Pinker will be on the nativist side of this debate.  Four reasons:
1.  Evolution.
2.  Creole languages.
3.  Brain structure.
4.  Patterns of language acquisition.
B.  Is language cognitively special?  Is there dedicated brain hardware for doing language?  Or is it just like anything else that you do?  Pinker's for dedicated hardware.  Two reasons:
1.  People can have normal intelligence, but impaired language.
2.  People can have impaired intelligence, but normal language.
This comes down to:  Are there language modules?  A module is an encapsulated processor that takes a specific input and generates a specific output.  It can't be influenced by any outside information.  The other side of this debate favors a more interactive process.  If outside knowledge can influence a process, then it's probably not modular.  If outside knowledge can't influence a process, then it probably is modular (example of Müller-Lyer illusion).
III.  Pinker's arguments.  Here are the four things that Pinker is pushing to make us believe language is innate.  These are in no particular order, read chapters 1, 2, 10, and 11 for the background.
A.  Evolution.  There are a bunch of arguments here.  I think the first one is that we shouldn't expect animals to have language (not even monkeys).  The reason some people think they should is because they don't understand evolution.
I think a more interesting discussion is this.  Creationists say that something as fully developed and as perfect as language couldn't have evolved.  For example, Paley says if you find a rock on the ground, it's just a rock and there's no need to question its origins.  But, if you find a watch, you want to know who the watchmaker was.  If we want to argue for evolution of language, we have to know what the mechanism was and what some first approximations to language might have been.  No matter where you stand, language couldn't have emerged complete in one generation.
Cosmides and Tooby suggest that a whole host of cognitive abilities (from face recognition to language) developed because we play a little game called the prisoner's dilemma.  The idea:  If I could get something good by ratting you out, and I could get in big trouble by not ratting you out, what's my incentive to cooperate with you?  Answer:  The more we play this game, the more likely we are to get what we need.  Imagine we both go hunting and I get meat and you don't.  If you say:  "Give me meat now, and next time I get some, I'll share with you", you're proposing this kind of game.  You could be planning to cheat.  But, if you always cheat, you won't eat, and you'll die.  So, there's incentive to cooperate.  Obviously, abstract language helps us describe what we're willing to offer in trade at a point in time removed from our own.
We can actually experience the forces at work in this situation if we play this prisoner's dilemma:
You have two players with two cards each (a 'C' card and a 'D' card).  On each turn, each player will (anonymously) flip one card.  Scoring:
Player 1 flips Player 2 flips Player 1 gets Player 2 gets
Card C 
Card C 
Card D 
Card D
Card C 
Card D 
Card C 
Card D
$5 (R) 
-$5 (S) 
$10 (T) 
$0 (P)
$5 (R) 
$10 (T) 
-$5 (S) 
$0 (P)
Look at the pattern that evolves.  To maximize rewards, you need to cheat a little, but you need to be smart about it.  You obviously can't always cheat or the other person will too.  You can't never cheat, or the other person will always cheat.  You can't be predictable.  It turns into that comedy routine from the Princess Bride when one cup was poisoned and one wasn't.  "I know you know I know you switched the cups, but you knew I'd know that, so maybe you didn't switch, but you'd know I'd think that so you might have..."
What does language do for you here?  You have to characterize biologically significant events in terms of costs and benefits and you have to be able to explain how you're bargaining for those things.  Explaining it takes language.  Even if you don't have full blown language, every little bit helps.  This might be a reason why language is good, but where does it come from?
Let's look at a sequence from Bickerton's Language and Species.  Bickerton goes through a long list of things that would be required and the sequence.  I'm sketching in the rough outline, but there's a lot more.
1.  You need representation.  You can start with cells that respond especially to prey and not much else (like frog cells that respond to small, black objects moving through the visual field).  But, as Bickerton puts it, the only thing you can be sure of if your model is "see prey -> eat it" is indigestion.  Instead, what if you could sum the inputs from two or more cells, and these could be wired to assess internal as well as external events.  With just two cells, one for sensing prey and one for sensing satiation, you can get four states and choose one of two actions that's appropriate.
+"belly full" and -"prey present" = "no action"
-"belly full" and -"prey present" = "no action"
+"belly full" and +"prey present" = "no action"
-"belly full" and +"prey present" = "eat"
In other words, only eat when there's something to eat and you're hungry.  This is a primitive form of mental representation.
2.  You need categorization.  You have to have at least 'x' and 'not x'.  You might start with "my species" and "not my species."  That helps with reproduction.  Then you can differentiate.  The more of this you can do, the better.
3.  You need to distinguish between classes and particulars (like common vs. proper nouns).  Monkeys can do this.  If they call something a "banana" and then they eat it, they know the word applied to a class of things and not that one banana.  On the other hand, Roger is always Roger.  If you eat Roger, he's all gone.
4.  You need emotion.  This could be a first step in developing consciousness.  For example, an antelope needs to be afraid of a lion to fully appreciate the need to carefully study lion behavior and know when you can ignore the lion and when you need to run away.
5.  You might need a big brain.  As humans evolved, brains got a lot bigger.  The causal direction is not clear.  Maybe big brains helped develop language, or maybe language grew the brain.
6.  You need a vocal tract.  Humans underwent the change that allowed this.
7.  It helps if you stand up.  This frees your hands to do a lot more, and that allows more representation stuff to go on.
8.  You need good vision.  This is required to jump from tree to tree (depth perception makes a difference).  Improving vision grows brains and improves planning and representation.
9.  Protolanguage provides benefits.  Any communication is good.  This is where the prisoner's dilemma comes in.  This might be the situation where gradually improving communication gradually improved survivability.
10.  The last step, from protolanguage to full language may have been a giant leap.  Around 100,000 years ago, people suddenly developed culture in the blink of an eye, and language probably came with that.  Maybe brains had finally reached the point where they were ready, and one little bit more pushed them over the edge.  Bickerton provides three sorts of evidence for this.
a.  Children.  In six months they can go from one word sentences ("fan") to "let's get a piece of rock and make it go ding."  There's a giant step.
b.  Creoles go from pidgen to language in one generation.
c.  People seem to have undergone a major transformation based on artifacts in the fossil record.
B.  Creole languages.  More from Bickerton.  The basic premise is this:  If there's an innate grammar, then it should emerge in situations where a native language is lacking.  Can we find situations like that?  You might think of looking at feral children (like wolf boys), but that won't work.  There's a critical period during which you need some linguistic input, or no full blown language can develop.
Instead, Bickerton looked for places where social forces might have deprived people of native language inputs.  Slavery and itinerant labor situations provided that.  Large groups of people were gathered together who spoke a variety of languages.  To communicate with each other and their masters, they generally adopted a small vocabulary of terms from the language of the people in charge (English colonies adopted English, French, French and so on).  The primitive copy of the dominant language is called a pidgen.  It's functional in the sense that you can get ideas across (like Tex-Mex Spanish), but it's not a language.  It lacks a grammar and most of the function words.  It's hard to get your ideas across in pidgins.  The interesting stuff happens when you look at the children of people who speak pidgins.  They speak a full blown language derived from the pidgen.  This language is called a creole.  Bickerton studied Hawaiian creole because it developed recently enough that pidgen speakers and the first generation of creole speakers were still around for Bickerton to study.  Here's the outline of the argument:
1.  Pidgen is not a language.  It's different for each individual who speaks it, communicating a lot of ideas is complicated, and almost impossible.  I have some pidgen examples.  Since it's not a language, if children learn a language based on the pidgen, they can't have learned it from the environment.
2.  Creole is a language.  It emerges full blown in the first generation of children born to pidgen speakers.  You might argue that all the creole borrows is a vocabulary.  It could borrow the grammar that makes it a language from one of the dominant languages in the area.  Bickerton says that's not possible.  Here are some of the reasons.
a.  It's grammatically different from the pidgen.
b.  It's grammatically different from the dominant language (in this case English).  Some examples: c.  It's grammatically different from the other languages present.  Even when it shares features, it has some from this language, some from that.  It's not possible that the kids could have gotten together and chosen which parts of each language to borrow.  People widely distributed in space with exposure to different stuff all developed the same language structure.
3.  Creoles are the same the world over.  Some similarities:
a.  Verb conjugation.
b.  Order of markers.  It always goes tense, modality, aspect.
c.  Accomplished vs. not accomplished is always marked.
4.  Children learning language appear to start with something like creole.  I have some examples of that.  If you look at language acquisition, there are some mistakes that appear to occur because they're part of creole, and kids have to be trained out of them.  Two of these:
a.  Double negatives.  Here's a conversation recorded by McNeill.  (C = child, M = mother.)
C:  Nobody don't like me.
M:  Nobody likes me.
C:  Nobody don't like me.
M:  Nobody likes me.
C:  Nobody don't like me.
M:  Nobody likes me.
C:  Nobody don't like me.
M:  Nobody likes me.
C:  Nobody don't like me.
M:  Nobody likes me.
C:  Nobody don't like me.
M:  Nobody likes me.
C:  Nobody don't like me.
M:  Nobody likes me.
C:  Nobody don't like me.
M:  Nobody likes me.
C:  Nobody don't likes me.  (exasperated)
This kind of formation is common in children learning English even though it's not a part of English.
b.  Only using intonation to mark a question.  This is common to all creoles.  Instead of saying "Can you fix it?" children say "You can fix it?" with rising intonation at the end.  Obviously, English doesn't routinely work like that.
Other kinds of mistakes could happen if basic principles of learning are all that are operating, but they don't.  Bickerton says they don't because they're naturally part of the creole kids are trying to learn.  Two examples:
a.  Using "-ing" to mark continuing action is learned very early ("I sitting high chair"), but overextensions do not happen.  For example, you cannot use "-ing" to mark the duration of statives like "like."  You don't see any kids say "I liking mommy" even though it's a reasonable mistake.  We can argue about adults saying things like "I'm liking this."  Maybe it's not a stative in that sentence?
b.  English makes use of specific vs. nonspecific reference.  I can say "John has never read a book" and follow it with "and he will never read a book."  I can't follow it with "and he will never read the book."  The first book was not referring to any particular book, so you can't start talking like there is some particular book later on.  On the other hand, if you say "John read a book yesterday" you can finish with "and he enjoyed the book" but not with "and he enjoyed a book."  Children as young as three could make these distinctions, even though this is a hard task.
Given this, Bickerton wants to argue that creoles reflect the operation of the language bioprogram.
C.  Brain stuff.  Let's take a quick tour of the brain.  You've probably all seen the homunculus in the sensory and motor cortex.  It's a map of how your body projects onto the brain.  Sensory cortex senses inputs, motor cortex produces motor movements.  Besides these two areas and a few others, there are not a lot of places in the brain that can be reliably mapped onto specific functions.
Here's an outline of parts that can be identified in the left hemisphere (not counting the lobes).  Note all the language areas.  The principle way of mapping language areas is finding people with deficits, imaging their brains, or autopsying them, and matching deficits to brain damage.  Language deficits are called aphasias.  These are disorders of language that are not caused by some specific motor impairment.  In other words, they're entirely brain based.
Here's an outline of the major vascularization of the brain.  There are approximately 200,000 aphasias that come about as a result of head trauma each year in the US.  There are around 100,000 aphasias that come about as a result of stroke.  Stroke aphasias are more informative because the parts of the brain that die are usually more specific.
Here are the language pathways.  For speaking a heard word, you go from primary auditory cortex to Wernicke's area to the arcuate fasciculus to Broca's area to motor cortex and then out.  For speaking a written word you go from primary visual area to the angular gyrus to Wernicke's area to the arcuate fasciculus to Broca's area to motor cortex and then out.  Damage to any of these will cause an aphasia.  Here are the seven types of aphasia (from Kandel, Schwartz, and Jessell, 1991).
1.  Wernicke's aphasia.  Damage to Wernicke's area.  Deficit in comprehension.  Big damage can impair both written and spoken language comprehension.  Spoken tends to go first.  Speech is fluent, but speakers tend to have incorrect word choices and lots of confabulation.  You get an abundance of words, but little meaning.  Not necessarily aware of the problem.
2.  Broca's aphasia.  Damage to Broca's area.  Comprehension is usually preserved, production is poor.  This is true of spoken and signed languages.  Most of the bits that make grammar are left out.  Speech is slow and halting.  Repetition is also impaired.  Broca's aphasics tend to be aware of the problem.  There's a lot of emotional baggage that comes with this.
3.  Conduction aphasia.  Damage to the arcuate fasciculus.  Fluent speech, but substitution errors.  Comprehension is OK.  Can't repeat.  Naming is impaired, writing can also be affected.
4.  Anomic aphasia.  Damage to areas near the temporal-occipital border.  Can't select words correctly.  Relatively rare.
5.  Global aphasia.  Damage to the whole perisylvan area including Broca's, Wernicke's, and the arcuate fasciculus.  Patients cannot speak or comprehend language.  They can't read, write, name objects, or repeat.
6.  Transcortical aphasia.  Several kinds.  Transcortical motor aphasia disconnects Broca's from motor cortex.  Patients cannot produce creative speech, but they can repeat and name objects.
7.  Subcortical aphasia.  Damage outside of cortex.
Using evidence from these patients, there seems to be a very consistent organization of brains and language.  The argument goes:  If brains always wire up the same way, it must be genetically determined, so language must be innate.
Here are some other brain reasons that language might be innate:
1.  Localization of language in the left hemisphere.  Even true of signers.
2.  Language is related to anatomical differences between the two hemispheres.  The language areas of the left are bigger.
3.  This asymmetry appears at 31 weeks gestation, so it can't be due to experience.
4.  Infants are born with really good ability to discriminate sounds.  This actually gets worse as they learn a particular language.
5.  There are a lot of regularities in acquisition.
6.  There's a critical period for learning language.
All together, they argue for some contribution of biology to language.
D.  Acquisition.  Pinker highlights the difficulty facing a child learning language with this example about the English agreement suffix "-s" in "he walks."  To use it right you have to know: To learn this you have to notice that verbs sometimes end in "-s," search for examples of this, and figure out how tense, aspect, person, and number all feed into it.  You also have to notice that number of syllables in the last word isn't important, etc.  This is so hard that kids shouldn't be able to do it.
There's also a pretty rigidly fixed sequence of language acquisition (see overhead).  Fixed sequences suggest maturation processes, maturation processes suggest biology, biology suggests innate language.
That's Pinker's case.  Now that we know it, let's look at the other side.  Watch Elizabeth Bates' speech.  Compare and contrast.  I apologize in advance for not trying to match up with Bates' order.

Langston, Psychology of Language, Notes 9 -- Development

Note:  These notes are in development.  Eventually, a new and nicer version will go here.

I.  Goals:
A.  A framework.
B.  Sequence of acquisition.
C.  An exercise.

II.  A framework:  As we think about language acquisition, I want you to consider how this relates to the nature/nurture debate.  Is language innate, or can it be learned?  This is one of the fundamental questions in psycholinguistics.
The format for this lecture will be:
A.  Major stage.
1.  Sub-stage.
a.  Comprehension/exposure (C/E):  What infants are being exposed to and how much of that they understand.
b.  Production (P):  What language stuff they can make.
c.  Communication (C):  What they can communicate with their language stuff.
If a category is left blank it indicates either that nothing happens at that stage or that nothing has changed since the previous stage.

III.  Sequence of acquisition.
A.  Prelinguistic.
1.  In the uterus: 
a.  Comprehension/exposure:  There’s evidence that infants are perceiving some speech sounds in the womb.  Newborns prefer their mother’s voice to other voices, and there’s even evidence that newborns recognize Dr. Seuss stories that were read to them in the womb.
b.  Production:
c.  Communication:
2.  Gestures (roughly through the first year).
a.  C/E:  Adult speech to infants is tailored to infant attention spans.  Some characteristics:
1)  Higher in pitch, more variable in pitch, more exaggerated intonation contour.  Newborns prefer baby-talk to adult speech, so it must work for them.
2)  Turn-taking:  Parents impose turn-taking from the start.  Any response from the infant (a burp, a fart, a giggle) is interpreted as a turn.
3)  Syntax:  Sentences are simpler (“This is a lion.  It’s a big lion.  His name is Leo...”).
b.  P:  Scream and cry for attention.  Early on, there’s no communicative intent.  Around 8 mos., you start to see intent develop.  How is that assessed?
1)  Wait for response.
2)  Persistence.
3)  Alternative plans if they’re unsuccessful.
Anecdotally, you can see evidence of intent by looking at what the infant is looking at.  If they’re trying to call attention to a dog and they’re looking at the dog, that probably doesn’t have intent.  But, if they’re looking at a person’s face to make sure the person is looking where they want them to look, that’s intent.
c.  C:  They can convey:
1)  Assertion:  Noting that something exists.
2)  Simple requests:  Point to what they want.
B.  Sounds.
1.  Cooing:  Practice with vowels.
a.  C/E:
b.  P:  They make back vowel sounds first.
c.  C:
2.  Babbling:  Making speechlike sounds.
a.  C/E:  Parents love this, so they’re exposed to a lot of positive feedback.  But, parental feedback plays a limited role in making this start or in keeping it going.  In terms of comprehension, they’re reorganizing phones into phonemes.  Until babbling, any infant from any language environment can recognize just about any speech sound.  After babbling, the sounds are mapped into a particular language.  So, English speaking infants lose the ability to distinguish the /k/’s in “keep” and “cool”, and Japanese infants lose the /l/-/r/ distinction.
b.  P:  Stages in babbling:
1)  Reduplication:  babababa.
2)  Variegated:  bapabapabapa.
3)  Complex:  digodapalaba.  At this point, English-speaking babies start to sound like English speakers, even though they’re not making real words.
c.  C:  Babbling isn’t about communication.  They babble alone, and parents can’t modify it with reinforcement or punishment.  Instead, it seems to be about learning the sound system.
3.  Between sounds and words:
a.  C/E:  The child might begin attending to morphology.  So, get beyond sounds, and look for meaningful units.
b.  P:  No sounds coming out.
c.  C:
C.  One-word stage:
1.  Idiomorphs:
a.  C/E:
b.  P:  Get a fully realized locutionary act.  The kid realizes that the sound stuff they’re making has meaning.  But, their words are idiosyncratic.  Their own parents probably know what they’re saying, other people probably won’t.
c.  C: 
2.  Holophrases:  Still one word, but the idea is kids have a grammatical utterance inside that can’t get out, only one word emerges.
a.  C/E:  Realize that things have names.  They focus on concrete objects, but on particular kinds of concrete objects: 
1)  They like things that move (doggie, truck).
2)  They like things they can act upon.
This is the beginning of the word explosion.  They learn an average of eight words/day from now until they’re six.
b.  P:  Difference between their words and adult speech:
1)  Reduction:  “bottle” = “baw.”
2)  Coalescence:  “pacifier” = “paf.”
3)  Assimilation:  “dance” = “nance.”
4)  Reduplication:  “daddy” = “dada.”
Each of these eases he burden on pronunciation.  But, is it a problem with production or comprehension?  Do they say it this way because that’s how they hear it?
You also get over- and under-extension.  Over-:  Apply a label too broadly (all four-legged animals are doggie).  Under-:  Too narrow (only Mom’s shoes are called shoe).
c.  C:  Holophrases are ambiguous.  If you want milk and say “milk” and Mom says “Yes, that’s milk”, the whole experience can be frustrating.  This is some incentive to move to the next stage.
D.  Two-word stage:  Called telegraphic speech because the child sounds like a person sending a telegraph.
a.  C/E: 
b.  P:  Producing two-word “sentences.”
c.  C:  They can express a much wider range of ideas:
1)  Nomination:  “that book.”
2)  Notice:  “hi book.”
3)  Recurrence:  “more milk.”
4)  Non-existence:  “allgone milk.”
5)  Attributive:  “big book.”
6)  Possessive:  “Adam book.”
7)  Locative:  “book chair.”
The problem is you still have ambiguity here.  “Mommy sock” could mean possessive (“Mommy’s sock”), locative (“Mommy’s wearing the sock”), etc.
E.  Morphology:  Inflections and little words come in.  Conceptualize age not in years but in Mean Length of Utterance (MLU):  How many words are in each sentence the child produces (on average).  Morphology begins to emerge around 2.5 MLU (approx. 3 yrs.).
1.  C/E:  Parents encourage this, but there still seems to be a limited role to reinforcement.
2.  P:  Sequence of development:
a.  Present progressive (“I driving”).
b.  Prepositions.
c.  Plurals.
You get over-extensions here too.  For example, kids will say “goed” for the past tense of “go” instead of “went”.  U-Shaped learning:  Get it right (“go-went”), overextend (“go-goed”), get it right (“go-went”).  Explaining why this sequence emerges is a new problem in language development.
3.  C:  Range of concepts they can express increases greatly.
F.  And so forth...  They move into higher order grammatical structures.  By five they’re just about adults, although some work is still going on.  Around six they get passive.  Between five and 10 they get the difference between “John is easy to see” and “John is eager to see.”  Around nine they get the difference between “John promised Bill to go” and “John told Bill to go” (before that they think Bill goes in both cases).

IV.  An exercise:  To firm up our analysis of what’s required for acquisition, think about the problems on the next page.  They go in order:
A.  Do children need language exposure to develop language?  If they get absolutely no exposure, what happens?  (Genie, the wild boy)
B.  What if they get exposure to language, but no grammar?  (Creole Languages)
C.  What’s the role of parental feedback (how much explicit teaching is required)?

Psychology of Language Notes 9
Will Langston

 Back to Langston's Psychology of Language Page